基于记忆的布尔博弈与网络自组织现象
Memory based Boolean game and self-organized phenomenon on networks
摘要：
我们研究了一个发生在正则环上的基于记忆的布尔博弈（MBBG），其中每个智能体根据其在记忆中记录的最后M个时间步的局部最优状态来行动，少数智能体得到奖励.引入一个介于0和1之间的自由参数p来表示智能体根据记忆做出决策的意愿强度。我们发现，在适当的意愿强度p下，MBBG系统可以自发地演化到比随机博弈更好的状态;而当p较大时，系统会出现羊群行为，降低系统收益。通过分析系统的动力学行为对记忆容量M的依赖关系，发现记忆容量越大，越有利于系统出现更好的性能状态，有效地抑制羊群行为，从而提高系统收益。考虑到长时间记忆的高成本，本文还讨论了增强记忆容量以抑制羊群行为的问题，并建议M=5是一个较好的选择。
英文摘要：
We study a memory-based Boolean game (MBBG) taking place on the regular ring, wherein each agent acts according to its local optimal states of the last M time steps recorded in memory, and the agents in the minority are rewarded. One free parameter p among 0 and 1 is introduced to denote the strength of the agents' willing to make a decision according to its memory. We find that, given proper willing strength p, the MBBG system can spontaneously evolve to a state of better performance than the random game; while for larger p, the herd behavior emerges which reduces the system profit. By analyzing the dependence of the system's dynamics on the memory capacity M, we find that a higher memory capacity favors the emergence of the better performance state, and effectively restrains the herd behavior, therefore increases the system profit. Considering the high cost of long-time memory, the enhancement of memory capacity for restraining the herd behavior is also discussed, and the M=5 is suggested to be one good choice.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
